Effects of distance to neighbors. Genes are divided into urban (green) and rural (blue) classes on the basis of their distance to the closest gene neighbor on either side (urban genes are the 25% with shortest distance). (A) A scatter plot of mRNA/cell versus distance to closest neighbor (min dist). Note that negative values correspond to sizes of overlap for genes that overlap their neighbors, and that the scale on the X-axis is logarithmic. (B) Cumulative distribution functions of ln(mRNA/cell) by class, which show depletion of highly expressed genes in the urban class. (C) A scatter plot of ln(mRNA/cell) versus sign(min_dist) ln(min_dist). This represents the log transformation for min_dist, accounting for negative values (see Methods). The plot reveals a positive association. (D) The added variable plot for ln(mRNA/cell) versus sign(min_dist) ln(min_dist) after correcting for ln(gene length). The added variable plot shows a slightly increased positive association, evidence that the relationship detected between expression and distance to closest neighbor is not a spurious byproduct of the relationship between expression and gene length (i.e., not a consequence of urban genes being longer than rural ones). Red and black curves represent lowess smooths on the actual data and on random permutations, respectively. (corr) Correlation coefficient. p-values are in parentheses. (Right table) Mean and median distances to closest neighbor for genes in the urban and rural classes. (Left table) Mean expression with its standard error, on the original, mRNA/cell, and logarithmic, ln(mRNA/cell), scale, for genes in the two classes. The mean expression level for the urban class is almost twofold lower than for the rural class.
